On 27/08/13 15:44, Frank Wall wrote:
Hi Laszlo,
On 27.08.2013 14:24, Laszlo Hornyak wrote:
If I understand correctly, this is what happens to your installation, the AddVmFromTemplate command goes to fail and engine tries to compensate, and upon successful compensation you get VM records removed from DB and storage cleaned up. just to be sure, please note, that the task actually completed successfully from my point of view. The new VM (direved from the template clone task) is fully usable and on the webadmin the task is shown as "completed".
So maybe at some point in time the engine does get the job status wrong and the problem begins...
I need to look into this but I think it should be safe to remove this record while the engine is down. Make sure to make a backup if you try! OK, I did the following test:
- created a new VM from my custom template (with cloned Disk) - waited for the job to complete - manually deleted the related DB entry from the table "async_tasks" - set host/node to maintenance - rebootet engine
Result: The VM did NOT disappear! :-)
I think somehow ovirt-engine does get the results of the job wrong, thus it does not recognize that the job completed *successfully* and does a retry after reboot (which fails horribly and leads to disappearing VMs).
Thanks - Frank
Hi, I had the same experience with the latest beta version, ovirt-engine-3.3.0-1.el6.noarch After create VM from template was initiated, two rows where inserted in the table "async_tasks" . After the creation finished successfully (and the VM could be used), one row was deleted from the table but one was left there. If you restart ovirt-engine then the VMs (created from template) where deleted. Their disks were not deleted. The left over row was also deleted. If you manually delete the left-over row from the table before you restart ovirt-engine, then the VM is not deleted. Is there a BZ for this to monitor it's fix progress? regards, Giannis